A predictive model for relapse in high-risk primary breast cancer patients treated with high-dose chemotherapy and autologous stem-cell transplant.

High-dose chemotherapy (HDCT) is currently under evaluation for high-risk primary breast cancer (HRPBC), defined by extensive axillary nodal involvement or inflammatory breast carcinoma. Phase II studies of HDCT for HRPBC show that 30-40% of patients eventually relapse. We retrospectively reviewed 176 patients enrolled in clinical trials of HDCT for HRPBC at the University of Colorado and analyzed 23 potential predictive variables for relapse. All of the patients received the same regimen, with cyclophosphamide, cisplatin, and BCNU. Nine patients who experienced a toxic death were excluded from this analysis. The resulting predictive model was subsequently tested in an independent patient set treated at Duke University with the same HDCT regimen. Nodal ratio (number of involved nodes:number of sampled nodes), tumor size, grade, stage, estrogen receptor, progesterone receptor, and clinical inflammatory breast carcinoma correlated with risk of relapse. Nodal ratio, tumor size, and the combined estrogen receptor/progesterone receptor status were independent predictors. A scoring system using those three variables determines the risk of relapse, with a sensitivity and specificity of 60 and 90%, respectively, and a positive and negative predictive value of 65 and 88%, respectively. The differences in relapse-free survival and overall survival between high- and low-score patients were highly significant (P<0.000001). This model was subsequently validated in the Duke patient set. This model can identify two subgroups of HRPBC patients with low (12%) and high (65%) risk for recurrence after HDCT. Future research that tests new therapies will focus on those patients with a high score.